backend/config.py

"""Runtime configuration defaults.

Only HF_TOKEN is intentionally left without a default because chat, embeddings,
and dataset writes need a Hugging Face token in deployed Spaces.
"""
from __future__ import annotations

import os
from pathlib import Path


REPO_ROOT = Path(__file__).parent.parent


def _default_dataset_id() -> str:
    space_id = os.environ.get("SPACE_ID") or os.environ.get("HF_SPACE_ID")
    if space_id and "/" in space_id:
        return f"{space_id}-corpus"
    return "iam-earth-dev-corpus"


def _env_bool(name: str, default: bool = False) -> bool:
    raw = os.environ.get(name)
    if raw is None:
        return default
    return raw.lower() in {"1", "true", "yes", "on"}


HF_TOKEN = os.environ.get("HF_TOKEN")
DATASET_ID = os.environ.get("DATASET_ID") or _default_dataset_id()
ADMIN_PASSCODE = os.environ.get("ADMIN_PASSCODE", "password")

LLM_MODEL = os.environ.get("LLM_MODEL", "Qwen/Qwen2.5-7B-Instruct")
LLM_MODEL_LOWER = LLM_MODEL.lower()
EMBED_MODEL = os.environ.get("EMBED_MODEL", "BAAI/bge-small-en-v1.5")
LLM_EXTRA_BODY_JSON = os.environ.get("LLM_EXTRA_BODY_JSON", "").strip()
LLM_FINAL_ANSWER_EXTRA_BODY_JSON = os.environ.get(
    "LLM_FINAL_ANSWER_EXTRA_BODY_JSON",
    "",
).strip()
SHOW_LLM_REASONING = _env_bool("SHOW_LLM_REASONING", False)

LLM_MAX_TOKENS = int(
    os.environ.get(
        "LLM_MAX_TOKENS",
        "2048" if "qwen/qwen3" in LLM_MODEL_LOWER or "qwen3" in LLM_MODEL_LOWER else "1024",
    )
)
LLM_FINAL_ANSWER_MAX_TOKENS = int(os.environ.get("LLM_FINAL_ANSWER_MAX_TOKENS", "768"))

BACKEND_PORT = os.environ.get("BACKEND_PORT", "8000")
FRONTEND_PORT = os.environ.get("FRONTEND_PORT", "7860")
BACKEND_INTERNAL_URL = os.environ.get("BACKEND_INTERNAL_URL", f"http://127.0.0.1:{BACKEND_PORT}")

TMP_UPLOAD_DIR = Path(os.environ.get("TMP_UPLOAD_DIR", str(REPO_ROOT / ".tmp_uploads")))
LANCEDB_DIR = Path(os.environ.get("LANCEDB_PATH", str(REPO_ROOT / ".lancedb")))
LANCEDB_PATH = str(LANCEDB_DIR)